Digital Elevation Models (DEMs) are currently produced by both manual and automated methods. Manual methods are typically reliable, but are slow and expensive for large areas. Automated methods, which determine the ground surface elevation by matching conjugate image portions, can be fast and relatively inexpensive but fail on complicated scenes and in featureless areas. Such automated techniques require the availability of powerful digital photogrammetric workstation with sophisticated software.
In this research, a semi-automatic procedure is presented to generate DEMs from stereo digital imagery. Here, the operator points to posts of interest in one image and their conjugate points are found, to sub-pixel accuracy, by use of matching. This would assure selecting appropriate matching entities, leading to a minimal number of matching ambiguities. Moreover, this procedure can be implemented on a PC which is always available in places that can not afford costly photogrammetric workstations. The test imagery consists of a stereo-pair of aerial images covering an urban area. The images are scanned with two different resolutions: 200 dpi and 600 dpi. A total of 90 feature points in the overlapping area are selected and matched using correlation technique. ِِThe 3-D ground coordinates of the selected points are computed using bundle adjustment with fixed and inner constraints. Prototype software is developed for matching and adjustment computations. The achieved results have shown simplicity and efficiency of the adopted procedure in reconstructing DEMs from digital aerial imagery.
|
